Research On Autonomous Localization Of Humanoid Robot In Indoor Environment

In the research field of humanoid robots,autonomous localization technology is the key premise for robots to perform complex tasks such as autonomous navigation and family care.The humanoid robot has complex structure and motion mode,which will cause great interference to sensor measurement.Therefore,it is of great significance to achieve accurate and stable autonomous localization.The main research work of this paper is to design and implement the autonomous localization method of humanoid robot in indoor environment.The specific research contents are as follows:Firstly,this paper uses a single sensor location method.The forward kinematics is used to locate the torso of the robot,and the true pose of the robot is estimated after coordinate system correction and scale restoration for Simultaneous Localization and Mapping(SLAM)based on monocular vision.The pose of kinematics solution and the Inertial Measurement Unit(IMU)carried by robots have high stability but large cumulative error.The monocular vision SLAM has high accuracy but depends heavily on the quality of feature points.Secondly,in order to achieve more accurate and stable autonomous localization of the robot,this paper designs a localization algorithm based on multi-information fusion,which combines the position of kinematics solution,the yaw measurement of IMU and the visual SLAM position estimation.Extended Kalman filter and unscented Kalman filter are used to compare and analyze the performance of multi-information fusion localization algorithm based on two different non-linear filters.The fusion results are applied to the footstep planning to correct the motion drift,and the closed-loop motion control based on multi-information fusion localization feedback is realized.Finally,a multi-information fusion localization system is implemented under ROS(Robot Operating System),and a series of experiments based on NAO humanoid robot are designed to verity the system in indoor environment.The experimental results show that,compared with the traditional kinematics solution method,the average error of the fusion multi-information localization method is reduced by 0.081 m indoors,the root mean square error is significantly reduced,the localization accuracy is effectively improved,and the localization can still be stably performed when the visual SLAM fails,which verifies the effectiveness of the proposed method.